Detecting Transient Surface Features with Dynamic Landmarking

Motivation

We have developed methods to dynamically and autonomously detecttransient surface features, such as dust devil tracks or dark slopestreaks on Mars, from images. Most prior work on this subject hasrelied on manual examination of image pairs. Exciting discoveriesof new surface features such as gullies and impact craters have beenmade, usually serendipitously. How many more such features remainundiscovered in the massive volume of images being collected andreturned?Automated methods can help reduce the manual effort needed to find andcatalog new and interesting features. Previous techniques forautomated analysis have focused on changes at the pixel level. Theyrequire an initial, sometimes slow, full registration between acandidate pair of images. Once the images are registered, subtractingone from the other yields changes. These are usually thresholded orsubjected to further analysis to help filter out noise andother uninteresting "changes".Dynamic Landmarking Approach

In contrast, our approach focuses on the image content, not just thepixels. We first analyze each image to identify visually salient"landmarks", and then compare the detected landmarks between theimages to highlight changes. No image registration is required. Weproceed as follows:

Detect landmarks: Using statistical measures of salience, create a"salience map" that indicates, for each pixel, how salient it is withrespect to its local region. Automatically select an appropriatesalience threshold, and use it to produce contours around regions ofhigh salience. Each such region becomes a "landmark." Landmarks mayinclude craters, volcanoes, fissures, and so on, as well as thetransient features we seek. Extract features: Compute descriptive attributes for eachlandmark, such as size (surface area), shape, albedo, homogeneity, etc. Using a trained landmark classifier, assign each discovered landmarkas one of a set of known classes of interest (craters, volcanoes, dustdevil tracks, gullies, etc.). Mark any unclassified landmarks as new,potentially high-interest regions. Create a Relative Landmark Graph (RLG): Landmarks (andtheir features) are the nodes, and edges connect each landmark to itsk nearest neighbors. Match the Graphs and Detect Changes: Compare the landmarksets discovered in two images to quickly detect any changes with highprecision. Using the Munkres/Hungarian algorithm, compute a matchingbetween the RLGs from two images. Mark unmatched landmarks as changes.

We have applied this approach to images collected by the Mars OrbiterCamera (MOC), Thermal Emission Imaging System (THEMIS), and otherMars-orbiting spacecraft.Benefits

• Because the landmarks are represented at an abstract level, it is possible to combine observations from different instruments at varying attitudes and under different illumination conditions.
  
• Dynamic landmarking can detect transient features without requiringimage registration, which represents a large step towards enabling theonboard use of this technology.
  
• These landmarks provide a regional characterization of the areacovered by the image that can be used to better understand surfaceprocesses as well as to recognize when two images overlap.
  
• One of the biggest benefits of this effort is the increasedproductivity it can lend to science investigations, as compared withmanual change detection. Our goal is to increase ourknowledge about other planetary surfaces, and the dynamic processespresent, to support future human and robotic exploration.